bioRxiv preprint doi: https://doi.org/10.1101/2020.05.30.124057; this version posted May 31, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Epigenome dysregulation resulting from NSD1 mutation in head and neck squamous cell carcinoma Nargess Farhangdoost1,2*, Cynthia Horth1,2*, Bo Hu1,2*, Eric Bareke1,2, Xiao Chen3, Yinglu Li3, Mariel Coradin4, Benjamin A. Garcia5, Chao Lu3, Jacek Majewski1,2$. 1 Department of Human Genetics, McGill University, Montreal, QC, H3A 1B1, Canada 2 McGill University Genome Centre, Montreal, QC, H3A 0G1, Canada 3 Department of Genetics and Development, Columbia University Irving Medical Center, New York, NY 10032, USA. 4 Biochemistry and Molecular Biophysics Graduate Group, University of Pennsylvania, Philadelphia, PA 19104, USA5 Department of Biochemistry and Biophysics, and Penn 5 Epigenetics Institute, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, 19104 USA * Those authors contributed equally to the manuscript $ Corresponding author [email protected] 1 bioRxiv preprint doi: https://doi.org/10.1101/2020.05.30.124057; this version posted May 31, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. ABSTRACT Epigenetic dysregulation has emerged as an important mechanism of oncogenesis. To develop targeted treatments, it is important to understand the epigenetic and transcriptomic consequences of mutations in epigenetic modifier genes. Recently, mutations in the histone methyltransferase gene NSD1 have been identified in a subset of head and neck squamous cell carcinomas (HNSCCs) – one of the most common and deadly cancers. Here, we use whole (epi)genome approaches and genome editing to dissect the downstream effects of loss of NSD1 in HNSCC. We demonstrate that NSD1 mutations are directly responsible for the loss of intergenic H3K36me2 domains, followed by loss of DNA methylation, and gain of H3K27me3 in the affected genomic regions. We further show that those regions are enriched in cis- regulatory elements and that subsequent loss of H3K27ac correlates with reduced expression of their target genes. Our analysis identifies genes and pathways affected by the loss of NSD1 and paves the way to further understanding the interplay among chromatin modifications in cancer. 2 bioRxiv preprint doi: https://doi.org/10.1101/2020.05.30.124057; this version posted May 31, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. INTRODUCTION Head and neck squamous cell carcinomas (HNSCCs) are very common and deadly cancers that can develop in oropharynx, hypopharynx, larynx, nasopharynx, and oral cavity1,2. These anatomically2- and genetically3- heterogeneous tumors can be induced either through some behavioral risk factors—such as tobacco smoking, excessive alcohol consumption, or insufficient oral hygiene4-7—or through human papillomavirus (HPV)8 and are currently classified into HPV(-) and HPV(+) groups9. HPV(-) tumors constitute around 80-95% of all HNSCCs10. The best currently available treatments have shown promising response in HPV(+) patients but have been less successful in HPV(-) cancers11-15. Thus, it is of great importance to understand the etiology of HPV(-) HNSCC tumors in order to develop more effective targeted therapies. Recently, mutations in epigenetic modifier genes, particularly the methyltransferase Nuclear Receptor Binding SET Domain Protein 1 (NSD1), have been implicated in HNSCC pathogenesis1,16. Subsequently, our group has identified H3K36M—Lysine to methionine mutations in histone H3 at the residue 36— mutations and demonstrated that NSD1 and H3K36M mutant HNSCCs form a distinct subgroup, characterized by specific DNA methylation (DNAme) patterns17. NSD1 is a histone lysine methyltransferase that catalyzes mono- and di-methylation of histone H3 at lysine 36 (H3K36me2)17-20. In addition, it may act as a transcriptional co-factor, responsible for activating or repressing different genes21. Mutations in other genes which encode H3K36 methyltransferases, such as NSD2 and SETD2, have not been frequently identified in HNSCC17, and it is not clear whether they contribute to this disease. Recent tumor-immune profiling of HNSCC patient samples has reported an association between NSD1 mutation and reduced immune infiltration20. In addition, it has been shown that HPV(-) tumors with NSD1 truncating mutations exhibit better treatment responses when targeted with cisplatin and carboplatin (chemotherapy based on platinum) compared to those lacking these mutations15. Thus, NSD1, its function, and the dysregulation it causes at the genetic and/or epigenetic level in HPV(-) HNSCC are of great importance for understanding the underlying mechanisms of tumorigenesis for improving the treatment responses. 3 bioRxiv preprint doi: https://doi.org/10.1101/2020.05.30.124057; this version posted May 31, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. We, and others, have further argued that the common mechanism underlying tumorigenicity in the H3K36me-dysregulated tumors is a reduction in H3K36me2 levels, followed by a global reduction in DNA methylation1,17,22,23. These observations, so far, have been based on primary tumor data, bulk quantification of epigenetic modifications, or data obtained from genetic manipulation of mouse embryonic stem cells17,24. Here, we use quantitative mass spectrometry of histone post-translational modifications (PTMs), genome- wide Chromatin Immunoprecipitation Sequencing (ChIP-Seq) and Whole Genome Bisulfite Sequencing (WGBS) to finely characterize the differences in epigenetic characteristics of NSD1-Wildtype (NSD1-WT) and NSD1-Mutant (NSD1-MT) HNSCC cell lines. We then use CRISPR-Cas9 genome editing to inactivate NSD1 in several independent cell lines and show that, in an isogenic context, the ablation of NSD1 is sufficient to recapitulate the epigenetic patterns that were observed in the patient-derived material. Furthermore, we perform RNA sequencing and thoroughly dissect the effect of the lack of NSD1 expression in HNSCC in order to link the epigenetic and functionality effects. We directly demonstrate the connection between NSD1, H3K36me2, Polycomb Repressive Complex 2 (PRC2)-mediated H3K27me3, and DNA methylation modifications in HNSCC. We also link the depletion of intergenic H3K36me2 with reduced activity of cis-regulatory elements– as profiled by the levels of H3K27ac – and reduced levels of expression of target genes. 4 bioRxiv preprint doi: https://doi.org/10.1101/2020.05.30.124057; this version posted May 31, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. RESULTS Epigenomic Characterization of NSD1 WT and Mutant HNSCC cell lines. We have previously shown that H3K36M and NSD1 mutations in HNSCCs are associated with low global levels of H3K36me217. More recently, we provided evidence that NSD1 mutant HNSCC samples are specifically characterized by low intergenic levels of H3K36me224. To confirm those results in a larger number of samples and characterize additional epigenetic marks, we examined three NSD1-WT (Cal27, FaDu, Detroit562) and three NSD1-MT (SKN3, SCC4, BICR78) patient-derived HNSCC cell lines. Mass spectrometry analysis demonstrates a clear difference in the global levels of H3K36me2 when comparing the mean of NSD1-WT with the mean of NSD1-MT samples (Fig. 1a, Supplementary Data 1). Visualization of H3K36me2 ChIP-seq tracks in representative regions illustrates that, in NSD1-MT cell lines, this mark is significantly reduced at the intergenic regions adjacent to genic regions (Fig. 1b). This intergenic depletion of H3K36me2 can be generalized to a genome-wide scale using heat maps and box plots (Fig. 1c, d). We note that there is significant variability across NSD1-WT cell lines with respect to the distribution of intergenic H3K36me2 (Fig. 1c): in FaDu, nearly all intergenic regions are marked with H3K36me2, and as a result intergenic levels are higher compared to genic, while Cal27 has the least pronounced intergenic H3K36me2 domains. This is further clarified with each cell line being illustrated individually (Supplementary Fig. 1). Those differences are likely to be an effect of cell of origin, presence of other oncogenic mutations, and relative activity levels of epigenetic modifier enzymes. However, our analysis shows a consistent and nearly total lack of intergenic H3K36me2 in all NSD1-MT cell lines, in contrast to genic levels that remains comparable to NSD1-WT lines. We have previously observed that NSD1 and H3K36M mutant HNSCC tumors are hypomethylated at the DNA level17 and, using mouse cell line models, proposed that this hypomethylation is mechanistically linked to the decrease in intergenic H3K36me2 via a reduced recruitment of the de novo DNA methyltransferase, DNMT3A24. Our extended analysis here clearly indicates that the decrease of intergenic H3K36me2, corresponds to a significant decrease in intergenic DNA methylation in all three NSD1-MT, 5 bioRxiv preprint doi: https://doi.org/10.1101/2020.05.30.124057; this version posted May